Fig. 9.

Bias of modeling accuracy (i.e., edge fidelity) to iPLV estimate and the construction of the IME A) iPLV estimates of true and false edges as a function of edge fidelity. Here, edges were pooled from 100 randomly chosen simulations (2×10⁴ true- and 8×10⁶ false-interactions); dashed lines indicate a range of threshold criteria ranging from 1.3 (relaxed) to 5 (strict). Above a given threshold green dots are true-positives (TP) and gray dots false-positives (FP), whereas below that threshold, green dots are false-negatives (FN) and gray dots true negatives (TN). B) Distribution of edge fidelity of true positive and false negative edges. C) The selection of the edge fidelity threshold for defining the edge fidelity mask M_fe (exclusion of edges estimated from poorly reconstructed loci). D) Tuning of the true negative rate. E) Visualization of the intractable-edge-mask in matrix form. The large dark area represents the deleted edges from poorly reconstructed loci (f_e<0.1), the fragmented dark areas within the heat map indicates the “short-range” edges that were deleted due to high local mixing (quantified by null hypothesis PLV₀).